On the HEAD branch, the following platforms have custom "init" targets
in MAKEDEV:

	amd64, i386, landisk, sbmips, sgimips

Why is this still necessary?  The comments seem to indicate a concern
about running out of inodes on the MFS mount if the default "all"
target is run.  However, the current MAKEDEV script counts the number
of inodes needed and ensures the created MFS mount is large enough to
hold all of the devices.  Shouldn't we just rely on the default "init"
target, which defaults to "all" on all of the other platforms?

I ask because I'd like to make it a easier to make a read-only-root
installation on my Xen boxes.

	Cheers,

	-- Johnny Lam <jlam@pkgsrc.org>